Management of power consumption in large computing clusters

Patent metadata

Field	Value
Publication number	US-2016147285-A1
Application number	US-201414553103-A
Country	US
Kind code	A1
Filing date	Nov 25, 2014
Priority date	Nov 25, 2014
Publication date	May 26, 2016
Grant date	—

Abstract

Management of power consumption in large computing clusters is disclosed herein. According to an aspect, a computing device comprising a power manager may be configured to receive, via a communication interface, information associated with the cluster of computing nodes. The power manager of the computing device may also be configured to determine whether a switch is coupled to an inactive computing node of the cluster of computing nodes based on the received information. Further, the power manager of the computing device may be configured to communicate a command to the switch to remove power supplied to a switch port of the switch coupled to the inactive computing node in response to determining that the switch is coupled to an inactive computing node of the cluster of computing nodes.

First claim

What is claimed:

1 . A computing device comprising:
a communication interface configured to communicate with a switch and a computing scheduler of a cluster of computing nodes; and a power manager comprising at least one processor and memory configured to:
receive, via the communication interface, information associated with the cluster of computing nodes;
determine whether the switch is coupled to an inactive computing node of the cluster of computing nodes based on the received information; and
in response to determining that the switch is coupled to an inactive computing node of the cluster of computing nodes, communicate a command to the switch to remove power supplied by a switch port of the switch coupled to the inactive computing node.

2 . The computing device of claim 1 , wherein the power manager is configured to receive, via the communication interface, the information associated with the cluster of computing nodes from the computing scheduler.

3 . The computing device of claim 1 , wherein the information identifies the inactive computing node in the cluster of computing nodes.

4 . The computing device of claim 3 , wherein the information identifying the inactive computing node comprises an address associated with the inactive computing node.

5 . The computing device of claim 4 , wherein the address comprises a media access control (MAC) address of the inactive computing node.

6 . The computing device of claim 1 , wherein the power manager is configured to determine whether the switch is coupled to the inactive computing node by comparing the received information to a configuration table associated with the power manager.

7 . The computing device of claim 6 , wherein the configuration table identifies the inactive computing node coupled to the switch port of the switch.

8 . The computing device of claim 7 , wherein the configuration table identifies the inactive computing node using an address associated with the inactive computing node.

9 . The computing device of claim 1 , wherein the command comprises information identifying the inactive computing node.

10 . A computing device comprising:
a communication interface configured to communicate with a plurality of switches and a computing scheduler of a cluster of computing nodes; and a power manager comprising at least one processor and memory configured to:
determine whether activity among the cluster of computing nodes is fragmented;
reallocate the activity among the cluster of computing nodes such that the activity amongst the cluster of computing nodes is not fragmented in response to determining that the activity among the cluster of computing nodes is fragmented;
determine whether a switch of the plurality of switches is not coupled to an active computing node of the cluster of computing nodes in response to the reallocation of activity amongst the cluster of computing nodes; and
communicate a command to the switch to power down in response to determining that the switch is not coupled to an active computing node of the cluster of computing nodes.

11 . The computing device of claim 10 , wherein the power manager is configured to:
receive, from the communication interface, information identifying inactive nodes within the cluster of computing nodes; and determine whether the activity among the cluster of computing nodes is fragmented based on the information.

12 . The computing device of claim 10 , wherein the power manager is further configured to receive, from the communication interface, information identifying inactive nodes within the cluster of computing nodes from the computing scheduler.

13 . The computing device of claim 12 , wherein the information indicates that the activity among the computing nodes in the cluster is fragmented.

14 . The computing device of claim 10 , wherein the power manager is configured to reallocate the activity among the cluster of computing nodes by communicating, to the computing scheduler, a command to defragment the activity among the computing nodes in the cluster.

15 . The computing device of claim 10 , wherein the power manager is configured to determine whether a switch among the plurality of switches is coupled to an active computing node of the cluster of computing nodes by accessing a configuration table associated with the power manager.

16 . The computing device of claim 15 , wherein the configuration table identifies the active and inactive computing nodes of the cluster of computing nodes coupled to each switch of the plurality of switches.

17 . The computing device of claim 15 , wherein the configuration table identifies that the switch of the plurality of switches is not coupled to an active computing node.

18 . A computing device comprising:
a communication interface configured to communicate with a cluster of computing nodes and a power manager of a plurality of switches coupled to each computing node of the cluster of computing nodes; and an activity scheduler comprising at least one processor and memory configured to:
receive, via the communication interface, information associated with the plurality of switches;
receive a request to schedule activity for an inactive computing node of the cluster of computing nodes;
determine whether one of a switch and a switch port of the switch coupled to the inactive computing node of the cluster of computing nodes is powered down based on the received information;
in response to determining the one of the switch and the switch port of the switch is powered down, communicate a command to the power manager to supply power to one of the switch and the switch port of the switch;
schedule activity for the inactive computing node to activate the inactive computing node.

19 . The computing device of claim 18 , wherein the information associated with the plurality of switches indicates that the one of the switch and the switch port of the switch is powered down.

20 . The computing device of claim 18 , wherein the command comprises information identifying the inactive computing node.

Assignees

• IBM

Inventors

• Alshinnawi Shareef F
• Cudak Gary D
• Hardee Christopher J
• Suffern Edward S
• Weber J Mark

Classifications

• G06F1/3287 Primary

  by switching off individual functional units in the computer system · CPC title

• G06F1/3206

  Monitoring of events, devices or parameters that trigger a change in power modality · CPC title

• G06F1/329

  by task scheduling · CPC title

• G06F1/3209

  Monitoring remote activity, e.g. over telephone lines or network connections · CPC title

• G06F13/4045

  where the bus bridge performs an extender function · CPC title